Temperature sensors
Acoustics Research Group
Peter T Gough
Visiting Professor
Acoustics Research Group,
Department of Electrical &
Computer Engineering,
University of Canterbury,
New Zealand
Contents
Acoustics Research Group
Why measure temperature?
Characteristics of interests
Types of temperature sensors
1. Thermistor
2. RTD Sensor
3. Thermocouple
4. Integrated Silicon Linear Sensor
Sensor Calibration (throughout)
Signal Conditioning Circuits (throughout)
E 80 Lecture 2
Why we need to measure temperature
Acoustics Research Group
Ideal gas law is PV = nRT
Sensors such as strain, pressure, force, flow, level,
and position many times require temperature
monitoring in order to insure accuracy.
As an example, pressure and force are often
sensed with resistive Wheatstone bridge
configurations. The temperature errors of the
resistive elements of these bridges can exceed the
actual measurement range of the sensor, making
the pressure sensors output fairly useless unless
the temperature of the bridge is known.
E 80 Lecture 3
Important Properties
Acoustics Research Group
Sensitivity
Temperature range
Accuracy
Repeatability
Relationship between measured quantity and
temperature
Linearity
Calibration
Response time
E 80 Lecture 4
Main single probe types
Acoustics Research Group
1. Thermistor
Ceramic-based: oxides of manganese, cobalt , nickel and
copper
2. Resistive Temperature Device -RTD
Metal-based : platinum, nickel or copper
3. Thermocouple
junction of two different metals
4. Integrated Silicon Linear Sensor
Si PN junction of a diode or bipolar transistor
E 80 Lecture 5
Acoustics Research Group
E 80 Lecture More than $10 6
Part I Thermistor
Acoustics Research Group
High sensitivity
Inexpensive
Reasonably accurate
Lead resistance ignored
Glass bead, disk or chip thermistor
Typically Negative Temperature Coefficient (NTC),
PTC also possible
R-T mode (zero-power mode):
nonlinear relationship between R and T
E 80 Lecture 7
Thermistor resistance vs
temperature
Acoustics Research Group
E 80 Lecture 8
Simple exponential Thermistor
model
Acoustics Research Group
RT = R0 x exp[ (1/T -1/T0)]
RT is the thermistor resistance ().
T is the thermistor temperature (K)
is a curve fitting parameter and itself is
temperature dependent.
Manufacturers will often give you R0, T0 and
an average value for
E 80 Lecture 9
Simple exponential Thermistor
Model
Acoustics Research Group
Usually T0 is room temp 25oC = 298.15oK
So R0 = R25
RT = R25 x exp[(1/T 1/298.15)]
where ln (R85/R25) /(1/358.15-1/298.15)
Not very accurate but easy to use
E 80 Lecture 10
Better Thermistor model
Acoustics Research Group
Resistance vs temperature is non-linear but can be
well characterised by a 3rd order polynomial; in this
case RT in terms of T.
ln RT = A + B / T +C / T2 + D / T3
where A,B,C,D are the characteristics of the
material used.
Total measurement uncertainty = +/- 0.005oC
E 80 Lecture 11
Inverting the equation: T in terms of R
Acoustics Research Group
The four term Steinhart-Hart equation
T = [A1 +B1 ln(RT/R0)+C1 ln2(RT/R0)+D1ln3(RT/R0)]-1
Also note:
A, B, C & D are not the same as A1, B1 , C1 & D1
Manufacturers should give you both for when R0 = R25
C1 is very small and sometime ignored (resulting in the three
term SH eqn)
E 80 Lecture 12
Aside that has nothing to do with
temperature sensors but is useful.
Acoustics Research Group
Both differentiating and integrating measured data is
difficult because of noise and unknown DC offsets.
Assume we have a 3rd order Voltage to Temp
relationship. i.e., V Vmeasured where
V = A + B T + C T2 + D T3
Then
dV/dT = B + 2 C T + 3 D T2
And
V dT = K + A T + 1/2 B T2 + 1/3 C T3 + 1/4 D T4
E 80 Lecture 13
Thermistor problems: self-heating
Acoustics Research Group
You need to pass a current through to measure the
voltage and calculate resistance.
Power is consumed by the thermistor and manifests
itself as heat inside the device
P = I2 RT
You need to know how much the temp increases
due to self heating by P so you need to be
given = the temperature rise for every watt of
heat generated.
E 80 Lecture 14
Heat flow
Acoustics Research Group
Very similar to Ohms law. The temperature
difference (increase or decrease) is related to the
power dissipated as heat and the thermal
resistance.
C=Wx
W in Watts
in oC /W
E 80 Lecture 15
Self heating calculation
Acoustics Research Group
oC = I2 RT Device to ambient
Example.
I = 5mA
RT = 4k
Device to ambient = 15 oC /W
oC = (5e-3)2 X 4e3 X 15 = 1.5 oC
This means the temp will read 1.5 degrees higher than ambient
Now try it for I = 1mA with RT = 15k
oC = ______________________________________
E 80 Lecture 16
Linearization techniques
Acoustics Research Group
Current through Thermistor is dominated by 10k
resistor.
E 80 Lecture 17
Linearization of a 1kOhm
Thermistor
Acoustics Research Group
This plot Ti = 50 0C, Ri = 275
E 80 Lecture 18
Linearization math
Acoustics Research Group
Vout/Vin = Ri/(Ri+ RT)
Find Ti from d(Vout/Vin)/dT =0 @Ti
then
RT = R0 x exp[/Ti /T0] x (- 2Ti)/(+2Ti)
where T0 , R0 and are given.
For example
T0 = 25oC (289 oK), R0 = 1 k, = 3560 oK
(NB. covert all temps to same scale)
E 80 Lecture 19
Linearization techniques
Acoustics Research Group
E 80 Lecture 20
Part II RTD
Acoustics Research Group
Resistive temperature devices (or detectors)
E 80 Lecture 21
pRTD, cRTD and nRTD
Acoustics Research Group
The most common is one made using platinum so
we use the acronym pRTD
Copper and nickel as also used but not as stable
E 80 Lecture 22
RTD are almost linear
Acoustics Research Group
RTDs have a + slope of relating the Resistance vs
Temp (i.e., resistance increases with temperature)
so that
RT = R0(1+ )(T T0)
Recognized standards for industrial platinum RTDs
are
IEC 6075 and ASTM E-1137 = 0.00385 //C
E 80 Lecture 23
Linearity: The reason RTDs are so popular
Acoustics Research Group
E 80 Lecture 24
Measuring the resistance
is easy using a constant current source
Acoustics Research Group
NB You can also use a simple voltage divider and
linearize it inside the ADC processor
E 80 Lecture 25
With long wires precision is a
problem
Acoustics Research Group
Two-wire circuits,
Three-wire circuits and
Four-wire circuits.
Most of the following circuits can be understood
using Ghms law (Goughs version of Ohms Law)
which I define as
No current = No voltage drop; or
No voltage drop = No current
E 80 Lecture 26
Two wire: lead resistances are a problem
Acoustics Research Group
Power supply connected here
No current flows in here
The IDAC block is a constant current sink
E 80 Lecture 27
Three wire with two current sinks
Acoustics Research Group
E 80 Lecture 28
Three wire with compensation
with two current sinks
Acoustics Research Group
E 80 Lecture 29
Four wire with one current sink.
Acoustics Research Group
E 80 Lecture 30
4 wire with precision current source
Acoustics Research Group
E 80 Lecture 31
Mathematical Modelling the RTD
Acoustics Research Group
The Callendar-Van Dusen equation
RT = R0 (1 + A T + B T2 + C T3(T-100) for T < 0 oC
= R0 (1+ A T + B T2) for T > 0 oC
where R0 is the resistance at T0 = 0 oC and
For platinum
A = 3.9083 x e-3 oC-1
B = -5.775 x e-7 oC-2
C = -4.183 x e-12 oC-4
E 80 Lecture 32
Experimentally
Acoustics Research Group
Mostly you need to derive temperature (+/-) from
the measured resistance (+/-).
Easiest way is to construct a Look-Up table inside
LabVIEW or your uP
Precision, accuracy, errors and uncertainties need
to be considered.
E 80 Lecture 33
Experimental uncertainties
Acoustics Research Group
For real precision, each sensor needs to be
calibrated at more than one temperature and any
modelling parameters refined by regression using a
least mean squares algorithm.
LabVIEW, MATLAB and Excel have these functions
The 0oC ice bath and the 100 oC boiling de-ionised
water (at sea level) are the two most convenient
standard temperatures.
E 80 Lecture 34
Part III Thermocouples
Acoustics Research Group
Type J thermocouple
E 80 Lecture 35
Thermocouples are very non-linear
Acoustics Research Group
E 80 Lecture 36
Type K and E
Acoustics Research Group
Type K (chromel {90% nickel and 10% chromium}
alumel {95% nickel, 2% manganese, 2% aluminium
and 1% silicon}) is the most common general
purpose thermocouple with a sensitivity of
approximately 41 V/C. Doesnt like high
temperatures
Type E (chromelconstantan) has a high output
(68 V/C) which makes it well suited
to cryogenic use. Additionally, it is non-magnetic.
Wide range is 50 to 740 C and Narrow range is
110 to 140 C.
E 80 Lecture 37
Type J and R
Acoustics Research Group
Type J (ironconstantan) has a more restricted
range than type K (40 to +750 C), but higher
sensitivity of about 55 V/C.
Type R (platinumrhodium alloy) containing 87%
platinum + 13% rhodium for the positive conductor
and pure platinum for the other conductor. Type R
thermocouples are used up to 1600 C.
E 80 Lecture 38
Type S and T
Acoustics Research Group
Type S (platinum-rhodium alloy) containing 90%
platinum + 10% rhodium for the positive conductor
and pure platinum for the other conductor. In
particular, type S is used as the standard of
calibration for the melting point
of gold (1064.43 C).
T (copper constantan) thermocouples are suited
for measurements in the 200 to 350 C range.
Since both conductors are non-magnetic, there is
no Curie point and thus no abrupt change in
characteristics. Type T sensitivity of about
43 V/C.
E 80 Lecture 39
Mathematical Model
Acoustics Research Group
To cover all types of thermocouples with one equation,
we need a 10th order polynomial to describe the
relationship between the voltage and the temperature
difference between the two junctions
Either
V = b0 +b1 x T + b2 x T2 +++++ b10 T10
+ bo exp(1(T-126.9686)2) for T>0oC
Or more usefully
T = a0 +a1 x V + a2 x V2 +++++ a10 V10
E 80 Lecture 40
Approximating noisy data with a
polynomial having an arbitrary
number of coefficients
Acoustics Research Group
% MATLAB script
% Assume we have measured at least 432 samples of the Voltage for monotonically increasing
% temperature T (say every one degree from 0 to 434 degrees C) and then fit a 10th order
polynomial
% Find the first 10 coefficients
number_of_coeffs = 10;
polycoeffs = polyfit([1:432], Vmeasured , number_of_coeffs)
% Since there is no semicolon, command above will print out all 10
% coefficents for you to check.
% Recreate the noise free approximation from the coefficients
Vsmooth = polyval(polycoeffs,[1:432]);
% Compare with the original data
Error = Vmeasured - Vsmooth;
%Now think about how you might use this concept to differentiate and to integrate
E 80 Lecture 41
10th order polynomial fit:
Find T from measured Voltage
Acoustics Research Group
E 80 Lecture 42
Filtering out the noise:
thermocouples are very noisy
Acoustics Research Group
E 80 Lecture 43
Look-up table with hardware compensation
is easier than using a polynomial
Acoustics Research Group
Note 8 bit accuracy
E 80 Lecture 44
What does 8 bit accuracy mean?
Acoustics Research Group
Eight bits = 28-1 levels = 255 levels
Assume supply voltage between 0 and 5 volts
Minimum V step between each level 20mV
Temp range required say 0 to 400 oC
Minimum temperature step 1.6 oC
i.e., Temp = T +/- 0.8oC
This determines the quantisation error of the result
regardless the accuracy of the sensor
However since the Temp often changes quite slowly you
can use averaging to increase the apparent accuracy
E 80 Lecture 45
Thermocouple
with compensation and filtering
Acoustics Research Group
Thermocouples are very noise prone & usually need filtering
Instrumentation amp
Low pass filter
E 80 Lecture 46
Using a uP to toggle between sensors
Acoustics Research Group
Useful when the reference junction is not close to the preamp & ADC
E 80 Lecture 47
Part IV: Si sensors
Acoustics Research Group
(semiconductor details for reference if ever needed)
The silicon bipolar transistor bandgap VBE
temperature sensor is an extremely common form
of temperature sensor used in electronic
equipment:
E 80 Lecture 48
The VBE as a function of T
Acoustics Research Group
VBE = VGO (1 T/T0) + VBEO (T/TO)
+ (nkT/q) ln(T0/T) + ( kT/q) ln(IC/IC0)
where
T = temperature in oK
T0 = reference temperature; often 300oK
VG0 = bandgap voltage at absolute zero
VBE0 = bandgap voltage at temperature T0 and current IC0
k = Boltzmann's constant
q = charge on an electron
n = a device-dependent constant
Change in VBE with Temperature --2mV/oC
E 80 Lecture 49
Matching transistors at different
temperatures
Acoustics Research Group
Difference in VBE = -2mV/oC
E 80 Lecture 50
Determining the
Voltage change wrt Temperature
Acoustics Research Group
By comparing the bandgap voltages of the same
transistor at two different currents, IG1 and IG2 and
where IG2 = N IG1, many of the variables in the VBE
equation can be eliminated, resulting in
VBE = (kT/q) ln (IG2/IG1)
= (kT/q) ln (N)
Usually this small voltage is multiplied by internal
amplifiers
E 80 Lecture 51
Switch the same transistor
between two different currents.
Acoustics Research Group
E 80 Lecture 52
Rats and Mice
(not covered in this lecture)
Acoustics Research Group
Noncontact IR single sensors.
Noncontact IR imaging cameras.
E 80 Lecture 53
So......References
Acoustics Research Group
Previous years E80 lectures on Temperature by Professors S.
Harris and Q. Yang
Wikipedia
Microchip Application Notes AN679, AN684, AN685, AN687,
AN871, AN893
Texas Instruments SBAA180
Baker, Bonnie, Designing with temperature sensors, part one:
sensor types, EDN, Sept 22, 2011, pg 22.
Baker, Bonnie, Designing with temperature sensors, part two:
thermistors, EDN, Oct 20, 2011, pg 24.
Baker, Bonnie, Designing with temperature sensors, part
three: RTDs, EDN, Nov 17, 2011, pg 24.
Baker, Bonnie, Designing with temperature sensors, part four:
thermocouples, EDN, Dec 15, 2011, pg 24.
E 80 Lecture 54
